In this talk, I will examine the computational motivations and empirical evidence for spatiotemporal dopamine (DA) waves that support reward learning within fronto-striatal networks. I will focus on the cognitive striatum as a case study to show that DA waves tailor decision signals according to local computational/behavioral specialty-- accomplished via vector-weighting delays in DA pulses across space and time. This code resolves key computational challenges in competing C-BG mixture of experts: spatiotemporal credit assessment at reward, and dynamic reprioritization of circuit inference and gating during performance. Ultimately, these DA wave dynamics represent an empirically informed revision of the longstanding "global broadcast" hypothesis of DA RPE signals. Finally, I will briefly summarize our recent attempts at understanding the complexity of the DA wave manifold, and competitive/collaborative circuit interactions that constrain DA to motif trajectories during specific task demands.